With recent technological advances in integrated circuit design, there are now as many as six to ten interconnect layers of a multilayer interconnect structure overlying the semiconductor transistors. Insulator layers separate the successive conductor layers. The conductor interconnect layers can have completely different conductor patterns and are connected to one another and to the transistor layer at different locations through vias extending vertically through the insulator layers. It is the formation of the vias with which the present invention is concerned.
To reduce resistive power losses in the integrated circuit, the interconnect layers and the vias typically employ aluminum and, more recently, copper as the principal conductor. The insulator layers are silicon dioxide, although more recently dielectric materials other than silicon dioxide having a lower dielectric constant are increasingly being employed. Because copper tends to diffuse over time through the insulator layer to cause catastrophic short circuiting, a barrier layer that blocks copper diffusion is placed between the copper material and the dielectric material wherever the two materials interface in the integrated circuit. The barrier layer is typically formed of an underlying tantalum nitride or titanium nitride layer contacting the insulator layer, an overlying pure (or nearly pure) tantalum (or titanium) layer and, finally, a copper seed layer over the pure tantalum (or titanium) layer. If the conductor is to be copper, then tantalum is preferred. The copper conductor is deposited on the copper seed layer. Such a barrier layer prevents migration or diffusion of copper atoms into the dielectric material. The tantalum and tantalum nitride (or titanium and titanium nitride) layers are poor conductors compared to copper. The formation of the barrier layer and of the copper conductors is carried out by physical vapor deposition. Other deposition processes may be employed for the metal fill step (copper deposition) such as chemical vapor deposition, plasma enhanced chemical vapor deposition or electroplating.
A problem arises in forming the vias extending vertically between the horizontal interconnect layers, as follows. Each vertical via opening is formed by etching a hole through an overlying horizontal insulator (silicon dioxide) layer so as to expose a portion of the copper conductor in the underlying horizontal interconnect layer. It is this exposed portion to which connection is to be made through the via. A barrier layer must be deposited on the interior surfaces of the via before the copper conductor can be formed in the via, to prevent copper migration as explained above. This barrier layer, in covering all interior surfaces of the via, covers the exposed portion of the underlying copper conductor. Since the barrier layer is an inferior conductor, it must be selectively removed from the underlying copper conductor (in an etch process) without removing the remainder of the barrier layer from the other interior surfaces of the via. This removal step has required interruption of the physical vapor deposition process in order to place the wafer in an etch chamber where the step of selectively removing the barrier layer from the underlying copper surface is carried out. The wafer is then returned to a physical vapor deposition reactor for formation of the copper conductor(s).
The interruption represented by the selective removal of the barrier layer entails a higher production cost and consumes production time. In recent years, a dual purpose reactor was developed capable of both physical vapor deposition of the barrier layer and selective removal of the barrier layer after the barrier layer formation step, without removing the wafer from the reactor. As a result, great savings in production cost and production time have been realized. This was accomplished by providing in the physical vapor deposition chamber a separate coil near the wafer. After barrier layer formation, the coil is used to form an inductively coupled plasma which selectively sputters the barrier layer from horizontal surfaces (i.e., the floor formed by the underlying copper conductor). Such selective sputtering (hereinafter referred to as “re-sputtering”) is achieved by applying RF bias power to the wafer to achieve an ion velocity distribution that is predominantly vertical. While this dual-purpose reactor works extremely well, it does entail some additional expense. For example, since the barrier layer deposition step involves sputtering a metal target and therefore deposits metal over all interior surfaces of the reactor chamber, the re-sputtering coil must be located inside the chamber so that no metallized surfaces shield the coil or otherwise prevent inductive coupling of RF power from the re-sputtering coil to the plasma. In order to avoid process contamination, the re-sputtering coil is formed of pure tantalum, adding cost. The coil is subject to very large temperature fluctuations during its lifetime, and must be changed periodically. RF power must be coupled to the re-sputtering coil through the vacuum seal of the reactor chamber and through an environment that periodically is completely filled with metal vapor. Therefore, an RF feedthrough must be employed that can tolerate the metal deposition, and whose exterior surfaces are textured to avoid excessive accumulation of deposited materials and flaking, and that can tolerate wide temperature excursions over its lifetime.
Another well-known dual-purpose reactor employs an external inductive coil overlying a portion of the ceiling not blocked by the metal sputter target. One problem is that the metal vapor deposition process can coat the ceiling with metal and thereby block inductive coupling from the coil antenna. A more significant problem is that the RF plasma produced by the coil produces a high proportion of metal ions from the target, so that the wafer bias cannot be optimized for etch selectivity to horizontal surfaces without impacting (de-optimizing) the flux of metal ions/vapor from the target. Therefore, the metal deposition process and the re-sputter process must be performed at separate times.
It should be noted that although such dual purpose reactors are capable of performing both the Ta/TaN barrier layer deposition step and the re-sputtering step, a different reactor is typically employed to perform the subsequent copper deposition step. This is because a high flux of copper ions on the wafer is desired, and the PVD reactor must be specially configured in order for the sputtered copper atoms to self-ionize in a very dense plasma at the copper target. Specifically, a very high D.C. power level (40-56 kWatts) is applied to the copper target and a specially configured magnetron is employed for a more concentrated plasma at the target. Because of the high density of copper ions near the target, it is placed very high above the wafer (390 mm), which limits the copper deposition rate to an acceptable threshold (as well as providing some beneficial collimation of copper neutrals). Typically, however, are large share of the copper ions are deposited on shields in the chamber and otherwise lost while traveling over this long distance.
In addition to the requirement for a copper PVD chamber and a barrier PVD chamber, a third chamber, an etch chamber, must be employed to carry out a pre-deposition cleaning process, since the copper PVD chamber and the barrier PVD/re-sputter chamber are not suitable for clean/etch processes.
Another problem is the tendency of the tantalum and/or tantalum nitride material deposited during formation of the barrier layer to deposit with non-uniform thickness along the via walls, and in particular to accumulate faster near the top corners of the vertical walls and thereby exhibit some tendency toward pinch-off. This makes it necessary to restrict the process window in order to ameliorate such problems. This problem is solved to some extent when, upon completion of the barrier layer deposition process, the re-sputtering process is performed, because the re-sputtering process tends to remove tantalum or tantalum nitride from the tops and corners of the via walls faster than elsewhere, while transferring tantalum (or tantalum nitride) material removed from the horizontal surfaces (floors) of the vias onto the lower portions of the via sidewalls. Nevertheless, it would be beneficial to avoid altogether the initial non-uniform tantalum or tantalum nitride deposition problem, to eliminate any risk of pinch-off, permitting some liberalization of the process window.
It would also be beneficial to avoid the necessity of the internal re-sputtering coil provided at least some of its benefits could be realized in a simpler manner.
The present invention provides benefits at least approaching those afforded by the internal re-sputtering coil without the need for such a coil. The present invention furthermore provides a way of ameliorating or avoiding non-uniform deposition of the barrier layer, and a way of avoiding or minimizing deposition of the barrier layer on the exposed copper conductor surface forming the floor of the via during formation of the barrier layer.
Another problem that is to be solved is that, as technological advances dictate smaller geometries and higher aspect ratios for vias and other features, the degree of ionization of the sputtered material (e.g., copper) must be increased to achieve the requisite conformality of the deposited film. Such an increase in ionization requires greater VHF power applied to the wafer support pedestal. Increased demands on process performance require that the temperature of the workpiece be more precisely controlled, dictating the use of an electrostatic chuck to clamp the wafer to a temperature-controlled surface. Use of an electrostatic chuck limits the amount of VHF power that can be applied to the wafer. This is, in large part, because the electrode to which RF power is applied in the electrostatic chuck typically consists of a small molybdenum mesh within an aluminum nitride puck. The small gauge of the mesh (e.g., 100 microns) greatly limits the efficiency of the mesh as an RF radiator and limits the amount of RF power that can be applied to the mesh to as low as 1.5 kW maximum. The requisite degree of ionization for conformal deposition on high aspect-ratio openings of small feature sizes (e.g., 45 nm) can only be achieved with much higher VHF power, e.g., 3.5 kW or higher.